BayesiaLab

BayesiaLab is a desktop application for Windows, Mac, and Unix that provides scientists with a comprehensive “laboratory” for machine learning, knowledge modeling, probabilistic reasoning (including diagnosis and simulation), causal inference, and optimization.

BayesiaLab utilizes the Bayesian network framework to gain deep insights into problem domains and reason about them under uncertainty.

BayesiaLab is the result of more than twenty years of research by Dr. Lionel Jouffe, Dr. Paul Munteanu, and their team of computer scientists. Their company, Bayesia S.A.S., is headquartered in Laval in northwestern France, with affiliates in the United States and Singapore. Today, Bayesia S.A.S. is the world’s leading supplier of Bayesian network software, serving hundreds of major corporations and research organizations around the world.

BayesiaLab in use for probabilistic modeling

Click to Zoom

Why BayesiaLab

One workflow, one platform: from model creation to analysis and optimization.
Human + data-driven modeling: build networks from expert knowledge, learn them from data, or combine both approaches.
Causal and predictive use cases: support diagnosis, forecasting, scenario analysis, and intervention planning.
Explainable outputs: graphical network structures and probability-based results help communicate findings.

Core Capabilities

Knowledge Modeling

BayesiaLab allows subject-matter experts to encode causal and probabilistic domain knowledge directly in a Bayesian network.
Nodes and arcs provide a direct graph representation of variables and relationships, with causal direction encoded through arc orientation.
Probabilistic relationships are represented through Conditional Probability Tables (CPT), with no fixed functional-form assumption.
Continuous variables can be discretized manually or automatically via workflow tools.
Expert elicitation is supported through the Bayesia Expert Knowledge Elicitation Environment (BEKEE).

Explore: Knowledge Modeling

Machine Learning

BayesiaLab provides algorithms for learning both structure and parameters from data.
Learning criteria are information-theoretic (for example, Minimum Description Length), and workflows do not require distributional assumptions.
Unsupervised Structural Learning discovers probabilistic structure without predefining input/output roles.
Supervised Learning targets predictive performance for selected targets while controlling model complexity.
Clustering Workflows include Data Clustering, Variable Clustering, and Multiple Clustering.

Explore: Machine Learning with BayesiaLab

Inference, Diagnosis, and Simulation

In BayesiaLab, diagnosis, prediction, and simulation are all evidence-conditioned inference workflows.
Bayesian networks represent a Joint Probability Distribution, enabling omnidirectional inference over all nodes.
BayesiaLab supports exact and approximate observational inference with hard, likelihood/virtual, probabilistic/soft, and numerical evidence.
Causal inference is available for intervention analysis (including Pearl’s Graph Surgery and Jouffe’s Likelihood Matching).
Effects can be analyzed through simulation-based workflows such as total effects and target mean analysis.

Explore: Inference: Diagnosis, Prediction, and Simulation

Model Utilization

BayesiaLab supports operational model use through adaptive and programmatic workflows.
The Adaptive Questionnaire selects the next best evidence to acquire by balancing information gain and acquisition cost.
WebSimulator publishes interactive models and adaptive questionnaires for web-based use.
Batch Inference and the optional Code Export Module support large-scale and embedded deployment.
The Bayesia Engine API provides modeling, inference, and learning capabilities through Java libraries.

Explore: Model Utilization

Knowledge Mining

Hellixia is BayesiaLab’s Generative AI assistant for turning questions, documents, and expert context into Bayesian networks.
The Automatic Causal Network Generator and Automatic Semantic Network Generator retrieve domain structure from prompts, including causal effects, CPT construction, and embedding-based semantic relationships.
Document Analysis transforms knowledge files into structured artifacts, including Semantic Flowcharts, Causal Semantic Diagrams, Knowledge Graphs, Semantic Networks, Causal Networks, and Doc-to-Node outputs.
Retrieval and enrichment workflows include Dimension Elicitor, Embedding Generator, Comment Generator, Class Description Generator, and Semantic Variable Clustering.
Causality Analysis and Causal Structural Priors retrieve and operationalize external causal knowledge as arc directions, explanations, and priors for downstream learning.
Node Translator and Image Generator support multilingual and visual enrichment of model model elements.

Explore: Hellixia

Knowledge Communication

BayesiaLab supports knowledge communication through graph-native representation, interactive simulation, and analysis.
Technical users can inspect model structure and behavior directly to extract interpretable insight from probabilistic models.
Visualization workflows include 2D, 3D, and VR-oriented perspectives for communicating relationships, effects, and scenarios.

Explore: Knowledge Communication